Enthusiasm for AI, particularly LLMs, may lead to self-deception, echoing Richard Feynman's warning about not fooling oneself. Despite tools feeling capable and enhancing perceived productivity, there's a lack of independent scientific evidence confirming increased productivity. A late 2025 study suggested developers using AI were actually 19% slower, even though they felt faster.

Anecdotal evidence and personal testimonies about AI-driven productivity gains are unreliable due to the difficulty of controlling for bias in personal experiments. Claims of AI causing mass layoffs are often "AI washing," masking typical economic corrections and over-hiring. Some companies are indeed laying off staff to fund unproven AI resources, driven by a fear of being left behind.

Furthermore, AI's economic sustainability is questionable. For instance, a $200 ChatGPT subscription starts losing money above 11% usage. The high operational costs suggest vendors may stop heavily subsidizing AI, drastically increasing its price and potentially diminishing perceived productivity benefits. With nearly a trillion dollars invested, the path to recouping these costs is unclear, especially given a bleak US economic outlook in August 2025.

The author, an AI skeptic influenced by Ed Zitron, believes AI's operational costs could lead to its discontinuation. Significant externalities like energy consumption, data center impacts, intellectual property theft, internet degradation, and memory hoarding raise ethical and moral concerns, making the author unwilling to use AI.